TL;DR: scratch filesystem on E5-26xx cluster id dead. It will be restarted from zero.
A silent corruption have made his way into the /scratch glusterfs filesystem of the E5-26xx cluster. The glusterfs filesystem is beyond repair.
Each attempt, since March 26th, to repair it has lead to a hang (kernel panic) on one or another node of the gluster's nodes.
The recent emergency poweroff, as the high temperatures (due to cooling system) in last months did not help.
As a consequence, the E5-26xx cluster will be restarted with an empty /scratch. All data has been lost (in /scratch/e5-2670-gfs-scratch).
(For Marco: /ssdscratch is OK. Breath.)
Lois, for PSMN Staff